When allocating assets, it is customary to consider the possibility that future returns may differ from those of the past and to adjust one’s forecasts accordingly. Such efforts occur less often with standard deviations and correlations. For the most part, even among institutions, what happened before is expected to happen again.
The estimates of asset-class correlations that underlie investment portfolios, even when made by highly sophisticated parties, tend to be created as follows: watch what transpires over several years, assume that pattern will continue, and be surprised - even shocked - if something occurs that has never occurred before.
This isn’t a criticism. It’s difficult to see how investment professionals could improve. Attempting to foresee all circumstances that might simultaneously torpedo US stocks, gold bullion, and mortgage bonds is almost impossible.
Worse, the financial markets are self-referential. Investment participants, however, are sentient, meaning that they react not only to what occurs outside of asset prices but also to changes within asset prices. They are susceptible to contagion. Even if analysts could accurately map every economic relationship between assets, an impossible task, their predictions would be derailed by investor behaviour.
Where investors do deserve blame, sometimes, is for taking their imperfect methods too seriously. It is one thing to recognise that, in the absence of a better approach, assuming that the future will resemble the past is a good starting point. Doing so outperforms the null assumption. It is quite another to treat correlations as if they measure something tangible and the investment task as if building a bridge.
Risk and Reward
Few if any investment professionals will confess to such a sin. They realise that their inputs are only estimates. And I believe them, to a point. They realise the assumptions that support their tools. However, when doing prolonged computations, as with asset-allocation optimisation software or value-at-risk assessments, it’s easy to forget the foundation’s instability. All too often, the answers to the calculations are regarded as solutions, not suggestions.
Reasonable minds can differ about the usefulness of quantitative routines that rely upon asset class correlations. They don’t appeal to me, but many Morningstar researchers disagree. What cannot be questioned is that mathematically derived asset allocations should also be appraised qualitatively.
In, particularly, their protections. Investors rarely fail to envision a portfolio’s rewards. The reason they invested in risky assets in the first place was the chance of achieving higher gains than if they held cash. In my experience, though, many do not consider how their portfolios might flop. What economic and/or financial-market conditions would be particularly harmful?
Conducting such exercises strikes me as more valuable than putting portfolios through quantitative tests. Once again, there is no escaping the past; as with hard-edged calculations, soft assessments begin with history. They, too, are imperfect. However, I find them to be more useful than black-box results because they provide a framework for thinking about adverse outcomes. They help with investment discipline.
Among the situations to be envisioned are: 1) recession; 2) an inflationary surge; and 3) widespread investor panic, perhaps due to market dysfunction. Those are merely brief examples. Outlining the various ways by which a portfolio might head south, all at once, is a topic for a future column. This article merely intends to persuade that such a thing can be likelier than one would have believed.
